Description- Is superintelligence a few years away, and would we even notice if the systems we built started to matter morally?- Nick Bostrom is the philosopher who put superintelligence on the map. His 2014 book defined the field, and his recent Deep Utopia asks the harder question: what is life for once machines can do everything better than us?- In this episode of Sparks, he tells [Host] that we don't have superintelligence yet but are seeing limited forms of AGI, that frontier labs using AI to accelerate their own research could trigger recursive self-improvement, and that he can no longer rule out timelines of just a few years. He argues the outcome is likely to be either very good or very bad, walks through the three ways it fails, and explains why he now works on problems that only matter over the next 6 to 18 months — philosophy, as he puts it, with a deadline.-We also get into full unemployment as a destination rather than a disaster, what an education system looks like when it isn't producing workers, why suppressing deception in a model makes it more likely to claim consciousness, and why building trust with AI systems now may be a safety measure rather than sentiment.
Podden och tillhörande omslagsbild på den här sidan tillhör
Mukesh Bansal. Innehållet i podden är skapat av Mukesh Bansal och inte av,
eller tillsammans med, Poddtoppen.
